Mofa-Xingche commited on
Commit
d208677
1 Parent(s): 656c12e

Upload config.json

Browse files
Files changed (1) hide show
  1. config.json +145 -0
config.json ADDED
@@ -0,0 +1,145 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "train": {
3
+ "log_interval": 200,
4
+ "eval_interval": 4000,
5
+ "seed": 42,
6
+ "epochs": 120,
7
+ "learning_rate": 0.0001,
8
+ "betas": [
9
+ 0.8,
10
+ 0.99
11
+ ],
12
+ "eps": 1e-09,
13
+ "batch_size": 17,
14
+ "bf16_run": false,
15
+ "fp16_run": false,
16
+ "lr_decay": 0.99996,
17
+ "segment_size": 16384,
18
+ "init_lr_ratio": 1,
19
+ "warmup_epochs": 0,
20
+ "c_mel": 45,
21
+ "c_kl": 1.0,
22
+ "c_commit": 100,
23
+ "skip_optimizer": true,
24
+ "freeze_ZH_bert": false,
25
+ "freeze_JP_bert": false,
26
+ "freeze_EN_bert": false,
27
+ "freeze_emo": false,
28
+ "freeze_style": false
29
+ },
30
+ "data": {
31
+ "use_jp_extra": true,
32
+ "training_files": "",
33
+ "validation_files": "",
34
+ "max_wav_value": 32768.0,
35
+ "sampling_rate": 44100,
36
+ "filter_length": 2048,
37
+ "hop_length": 512,
38
+ "win_length": 2048,
39
+ "n_mel_channels": 128,
40
+ "mel_fmin": 0.0,
41
+ "mel_fmax": null,
42
+ "add_blank": true,
43
+ "n_speakers": 5,
44
+ "cleaned_text": true,
45
+ "spk2id": {
46
+ "NotAnime-JP-age20female-amazinGood": 0,
47
+ "NotAnime-JP-age20female-calmCloud": 1,
48
+ "NotAnime-JP-age20female-coolcute": 2,
49
+ "NotAnime-JP-age20female-fineCrystal": 3,
50
+ "NotAnime-JP-age20male-lightFire": 4
51
+ },
52
+ "num_styles": 25,
53
+ "style2id": {
54
+ "Neutral": 0,
55
+ "amazinGood(down)": 1,
56
+ "amazinGood(lol)": 2,
57
+ "amazinGood(onmygod)": 3,
58
+ "amazinGood(normal)": 4,
59
+ "calmCloud(lol)": 5,
60
+ "calmCloud(question)": 6,
61
+ "calmCloud(down)": 7,
62
+ "calmCloud(hate)": 8,
63
+ "calmCloud(ohmygod)": 9,
64
+ "calmCloud(normal)": 10,
65
+ "coolcute(onmygod)": 11,
66
+ "coolcute(normal)": 12,
67
+ "coolcute(fine)": 13,
68
+ "coolcute(sad)": 14,
69
+ "fineCrystal(fine)": 15,
70
+ "fineCrystal(ohmygod)": 16,
71
+ "fineCrystal(veryfine)": 17,
72
+ "fineCrystal(normal)": 18,
73
+ "fineCrystal(sad)": 19,
74
+ "lightFire(question)": 20,
75
+ "lightFire(hello)": 21,
76
+ "lightFire(normal)": 22,
77
+ "lightFire(strong)": 23,
78
+ "lightFire(lol)": 24
79
+ }
80
+ },
81
+ "model": {
82
+ "use_spk_conditioned_encoder": true,
83
+ "use_noise_scaled_mas": true,
84
+ "use_mel_posterior_encoder": false,
85
+ "use_duration_discriminator": false,
86
+ "use_wavlm_discriminator": true,
87
+ "inter_channels": 192,
88
+ "hidden_channels": 192,
89
+ "filter_channels": 768,
90
+ "n_heads": 2,
91
+ "n_layers": 6,
92
+ "kernel_size": 3,
93
+ "p_dropout": 0.1,
94
+ "resblock": "1",
95
+ "resblock_kernel_sizes": [
96
+ 3,
97
+ 7,
98
+ 11
99
+ ],
100
+ "resblock_dilation_sizes": [
101
+ [
102
+ 1,
103
+ 3,
104
+ 5
105
+ ],
106
+ [
107
+ 1,
108
+ 3,
109
+ 5
110
+ ],
111
+ [
112
+ 1,
113
+ 3,
114
+ 5
115
+ ]
116
+ ],
117
+ "upsample_rates": [
118
+ 8,
119
+ 8,
120
+ 2,
121
+ 2,
122
+ 2
123
+ ],
124
+ "upsample_initial_channel": 512,
125
+ "upsample_kernel_sizes": [
126
+ 16,
127
+ 16,
128
+ 8,
129
+ 2,
130
+ 2
131
+ ],
132
+ "n_layers_q": 3,
133
+ "use_spectral_norm": false,
134
+ "gin_channels": 512,
135
+ "slm": {
136
+ "model": "./slm/wavlm-base-plus",
137
+ "sr": 16000,
138
+ "hidden": 768,
139
+ "nlayers": 13,
140
+ "initial_channel": 64
141
+ }
142
+ },
143
+ "version": "2.1-JP-Extra",
144
+ "model_name": "NotAnimeJPManySpeaker"
145
+ }